Radeon 8500 video overlay gone (xawtv and kdetv broken)

[Alex Deucher]

On Sun, Jan 2, 2011 at 11:47 PM, Robert Braddock
<robert at concordantthought.com> wrote:
> On Sun, Jan 02, 2011 at 07:45:16PM -0500, Alex Deucher wrote:
>> The radeon video input features are not supported with KMS.  If you
>> want to use them, you will need to disable KMS.  Boot with
>> radeon.modeset=0 on your kernel command line in grub.
>
> Partial success.  With that boot option, the device is there again
> (which is weird, because I had actually tried something else that was
> supposed to disable the KMS, and there wasn't any change that way), but
> it still doesn't actually work properly.
>
> Kdetv just shows some red lines (no surprise -- that's as much as it did
> before), and even though xawtv can put up the video, it will only show
> the video through a viewport sized like its default initial window
> (which is way too small).  If I resize or fullscreen the window, the
> image is scaled, but everything outside that default initial window size
> and position is black.  I say "default initial window" because even if I
> use a command line option to fullscreen it from the start, the visible
> portion is still that same tiny "window".  I'd really like to get it
> back to the workable state it was in before where I could at least
> resize it (even if I did have to bring up and close the menu to get it
> to redraw after resizing).
>
> I've tried various -this and -nothis options with no apparent change.
>

I'm not too familiar with the gatos code and I've never tested it
myself.  Best bet would be to bisect and try and figure out what broke
it.

>> I don't know of anyone actively working on porting the old video input
>> code to KMS.
>
> Can we get a warning message at least?  That would have really helped,
> especially if I'd still gotten a warning when the KMS was supposed to
> have been disabled.  I'd have known to try something else along that
> line, at least.

Sure.

>
> Is there any way to get a /dev/video0 from this kind of card/device?  I
> was optimistic that when X rolled the GATOS project in, the video-in
> support would finally be something resembling stable (I've completely
> given up on capture or tv-tuner functionality), but it looks like it's
> as tenouous and broken-by-default as ever.
>
> If the video-in code were "ported to KMS" would it work better?  I've
> never written video/driver code or worked inside of X, but I can do C.
> If this is just cosmetic stuff and something says what needs to be done,
> I could maybe do something (although X is definitely in the class of
> things I hoped never to be compiling from source again).
>
> But if X is just going to write the feature off anyway, I'd rather see
> it work through /dev/video so that I could at least use a different
> viewer program than xawtv.

KMS is really the only way to provide a v4l compatible interface for
the capture interface on the radeon cards.  You really need a unified
memory manager.  That's partly why the gatos code exposed the hw via
the current Xv hacks; the X driver was responsible for memory
allocation with ums which makes it's hard to interact with a v4l
kernel driver.

Most of the infrastructural work is done for porting the code to KMS.
The KMS driver exposes the i2c buses used by the various components
(tuners, decoders, etc.), so you should be able to use the existing
kernel v4l drivers for those components.  The hardest part would be
porting support for the VIP bus and the rage theatre asics to KMS.
All that's left after that is adding the probe code to the KMS driver
to load the proper modules (basically parse the multimedia vbios
tables) and registering a proper v4l interface.

Alex

>
> --
> Robert
>